Myrmecological News 19 171-183 Vienna, January 2014

Myrmica martini sp.n. – a cryptic species of the scabrinodis species complex (: Formicidae) revealed by geometric morphometrics and nest-centroid clustering

Bernhard SEIFERT, Ali BAGHERIAN YAZDI & Roland SCHULTZ

Abstract

Palaearctic populations of Myrmica so far known under the name M. scabrinodis NYLANDER, 1846 were studied by combining geometric morphometrics (GM) with nest-centroid (NC) clustering and hypothesis-driven data analysis. A new cryptic species, Myrmica martini sp.n., showing a rather limited geographical range extending over largely the montane to subalpine zones of the Pyrenees and French Alps, was identified. 41 landmarks and 252 semilandmarks were fixed in the clypeus, head capsule, mesosoma and petiole aspects of 359 workers belonging to 106 nest samples. Extracting the 14 most diagnostic shape components from a set of 316 relative warps and running these data in NC- clustering, resulted in a complete species separation despite minute interspecific differences and large overlap in any character. The species identification provided by NC-Ward, NC-K-means and NMDS-K-Means clustering and by the controlling linear discriminant analysis agreed in each nest sample. There was no classification in disagreement with zoogeographic data. The lectotype samples of the five most similar and possibly synonymous taxa had near-to-zero pro- babilities of belonging to the M. martini sp.n. cluster: M. scabrinodis (p = 0.00015), M. scabrinodis var. rugulosoides FOREL, 1915 (p = 0.0037), M. pilosiscapus BONDROIT, 1920 (p = 0.0008), M. sabuleti var. spinosior SANTSCHI, 1931 (p = 0.0006), and M. rolandi var. reticulata STÄRCKE, 1942 (p = 0.00001). Myrmica rolandi var. reticulata, of which a lecto- type was designated here, is established as junior synonym of M. spinosior whereas M. scabrinodis var. rugulosoides and M. pilosiscapus are confirmed as junior synonyms of M. scabrinodis. We provide a rather simple system to discri- minate M. martini from M. scabrinodis requiring 8 - 10 minutes of investigation time per specimen and resulting in an error of 3.6% on the nest sample level. Key words: , cryptic species, species delimitation, numeric morphology-based alpha-taxonomy, lectotype designation. Myrmecol. News 19: 171-183 ISSN 1994-4136 (print), ISSN 1997-3500 (online) Received 10 June 2013; revision received 21 October 2013; accepted 22 October 2013 Subject Editor: Christian Rabeling Bernhard Seifert (contact author), Ali Bagherian Yazdi & Roland Schultz, Senckenberg Museum für Naturkunde Görlitz, Am Museum 1, 02826 Görlitz, Germany. E-mail: [email protected]; [email protected]; [email protected]

Introduction Ants of the Myrmica LATREILLE, 1804 are widely In personal contacts with colleagues, the first author distributed across the Holarctic and have important func- frequently experienced criticism that such complex NU- tions in temperate ecosystems. The current estimate of the MOBAT approaches are unacceptable because of being Palaearctic fauna approaches 150 species (RADCHENKO & too time-consuming. Those critics confuse origin and end. ELMES 2010). The species delimitation within Myrmica First, we have to do our best in credibly delimiting species is poorly understood due to the overwhelming dominance on the basis of testable data sets under the best currently of subjective eye-inspection taxonomy instead of using available investigation standards. Then, and always after explorative and hypothesis-driven analyses of reproduc- this, we have to do the service to the public and to our- ibly recorded data. The crucial importance of numeric selves: finding the least time consuming and most simple morphology-based alpha-taxonomy (NUMOBAT) in the way to identify the recognized entities with an acceptable identification of cryptic species was emphasized by SEI- error rate. FERT (2009) who defined cryptic species as "two or more The identification of cryptic species in Myrmica is in its species which are not safely separable by primary visual or early beginning. NUMOBAT approaches delimited cryp- acoustic perception of an expert". This reflects the imme- tic species in the M. sabuleti, M. lobicornis and M. speci- diate sense of the word and restricts the terminus to the oides species complexes resulting in four rank elevations truly cryptic cases – i.e., to species not safely separable (SEIFERT 2000, 2005, SEIFERT & al. 2009) but also in four by training of innate pathways of the human cognitive synonymizations (SEIFERT 2011). This means a zero balance system. of accepted species numbers and rebuts the wide-spread opinion NUMOBAT could result in an oversplitting. Con- CLM and GM: Austria: Gnadenwald-2.1 km NE, 2012. sidering the continued philosophy of most taxonomists of 07.27, samples 1, 2 [47.337° N, 11.571° E, 1632 m]; Inns- prolifically publishing one new name after the other on the bruck: Seegrube, 2012.07.24, samples 1, 2 [47.305° N, basis of untested subjective impressions, it is not really 11.377° E, 1925 m]. – Czech Republic: Roudnicka, surprising that these detailed NUMOBAT analyses could 2012.09.24 [50.174° N, 15.829° E, 236 m]. – France: not identify a single undescribed taxon so far. Frasne, 2010.08.22 [46.831° N, 6.154° E, 840 m]; Jura: The methodology of NUMOBAT in ants experienced a Mouthe, lectotype of M. pilosiscapus [48.710° N, 6.190° E, fast evolution during the last years. In Myrmica, it started 940 m]; Tourettes-sur-Loup, 1955.06 [43.720° N, 7.060° E, with hypothesis-driven analyses of conventional linear mor- 400 m]; Tourettes-s-Loup, 1955.05 - 2 [43.720° N, 7.060° phometrics (SEIFERT 2000, 2005). Later, explorative ana- E, 400 m]. – Georgia: Borisacho, 1985.08.12, sample lyses of more complex data sets (SEIFERT & al. 2009, No 1197 [42.533° N, 44.933° E, 1500 m]; Diklo, 1985. SEIFERT 2011) and geometric morphometrics (BAGHERIAN 08.02, sample No 648u [42.402° N, 45.685° E, 1800 m]; YAZDI & al. 2012) were added to the toolbox. The most re- Omalo, 1985.07.31, sample No 1199 [42.380° N, 45.630° cent methodological improvement is NC-clustering (SEIFERT E, 1500 m]; Schatili, 1985.08.14, samples No 1195, 1196, & al. 2014). NC clustering (nest-centroid clustering) is a high- 632u [42.658° N, 45.159° E, 1450 m]; Shatili-Kuhweide, resolution explorative data analysis applicable to any group 1985.08.14, samples No 632, 633, 1197 [42.658° N, 45.159° of eusocial organisms or to any cohesive biological sys- E, 1450 m]. – Germany: Baindt: Kiesgrube, 2006.09. tem providing repeats of definitely conspecific elements. 12, sample No ANA2 [46.836° N, 9.664° E, 495 m]; Ba- Here, we provide the first case for Myrmica in which ruth, 1965.08.27 [51.230° N, 14.590° E, 150 m]; Baruth: advanced NUMOBAT resulted in the description of a new Schafberg, 2010.08.05, samples No 2, 3, 4 [51.232° N, cryptic species. We combined geometric morphometrics 14.601° E, 165 m]; Baruth: Schafberg, 2010.05.09, samples with NC-clustering and hypothesis-driven data analysis for No 782, 915 [51.232° N, 14.601° E, 165 m]; Baruth: Schaf- studying so far inseparable populations of the M. scabri- berg, 2010.05.23, [51.231° N, 14.601° E, 171 m]; Berz- nodis species complex and identified a new species. This dorf: Hutberg, 1982, sample No 819 [51.055° N, 14.886° E, new species, Myrmica martini sp.n., has a rather limited 280 m]; Bodensee: Gundholzen, 2005.10.08, sample No geographical range extending over largely the montane to 240 [47.699° N, 8.963° E, 443 m]; Breitlohmüs, 1990.05.05 subalpine zones of the Pyrenees and French Alps. It is [48.695° N, 8.419° E, 970 m]; Dubringer Moor, 1982.08.09 closely related to NYLANDER, 1846. / 12, sample No 1 [51.398° N, 14.163° E, 135m]; Engs- We firstly present the formal description of the new spe- tingen, 2012.09.10, sample No GK458 [48.393° N, 9.258° cies. Then we present our argumentation why Myrmica E, 782 m]; Grambach: Karlstadt, 1991.05.03, sample No martini sp.n. is no junior synonym of Myrmica scabrino- 225 [50.006° N, 9.774° E, 280 m]; Görlitz: Königshain-1 km dis var. rugulosoides FOREL, 1915, Myrmica pilosiscapus N, 1989.08.01 [51.198° N, 14.851° E, 300 m]; Heldrun- BONDROIT, 1920, var. spinosior SANT- gen, 1981.09.08 [51.300° N, 11.190° E, 130 m]; Hinter- SCHI, 1931, or Myrmica rolandi var. reticulata STÄRCKE, zarten-E, 2005.06.21 - 234 [47.908° N, 8.110° E, 880 m]; 1942. These taxa show a similar morphology and real or Hinterzarten-S, 2005.11.02 sample No 396 [47.983° N, potential geographic overlap. The type localities of M. spi- 8.105° E, 940 m]; Kunnersdorf-1 km SE, 1983.05.21 nosior and M. reticulata are within the range of M. mar- [51.197° N, 14.947° E, 258 m]; Lenzkirch-W, 2005. tini sp.n. and those of M. scabrinodis var. rugulosoides and 09.20 - 122 [47.865° N, 8.170° E, 844 m]; Löbauer Berg, M. pilosiscapus are 150 km north-northeast and 200 km 1983.05.10-Basalt [51.090° N, 14.693° E, 420 m]; Lö- north of the next known site of M. martini sp.n. bauerBerg: S-Rand, 1983.05.10, sample No N1 + 2 [51.088° N, 14.699° E, 360 m]; Kreba: Spisk, 1989.08.06 Material [51.361° N, 14.714° E, 144 m]; NSG Dellenhäule, 2006. Collecting dates are given in alphanumeric format yyyy. 05.21 / 22, samples No 1 - 6, 8 [48.816° N, 10.241° E, mm.dd. Sample numbers refer to the primary numbers in 582 m]; Oppitz: Initialmoor, 2003.07.15, samples No 1 - 3 the field books. [51.296° N, 14.423° E, 131 m]; Reinheim-Zeilhard, 1991. 06.02 - 035 [49.850° N, 8.780° E, 230 m]; Reichental-2 km Myrmica martini sp.n. ENE, 1990.05.05 [48.735° N, 8.415° E, 660 m]; Schmölln: A total of 23 nest samples with 84 workers was investigated FND Roter Berg, 1982 [50.930° N, 12.450° E, 230 m]; both by geometric morphometrics (GM) and conventional Sondershausen, 1983.05.29 [51.374° N, 10.834° E, 356 m]; linear morphometrics (CLM): Andorra: Andorra City- Utzenfeld-N, 2005.07.28 - 92 [47.841° N, 7.863° E, 1100 m]; 9 km NNW, 1991.05.17, samples No 5, 19, 25, 50, 77 Wolmatinger Ried, 2006.07.17, sample No 6673WOL [42.680° N, 1.470° E, 1850 m]. – France: Briancon, [47.674° N, 9.143° E, 393 m]; Wolmatinger Ried, 2006.11. 1955.06 [44.90° N, 6.63° E, 1300 m]; Fontainebleau, 10, sample No 7776WOL [47.703° N, 9.103° E, 395 m]; 1955.06 [43.33° N, 5.537° E, 500 m]; Saint-Martin- Würzburg: Dürrbach, 1991.05.27 - 068 [49.816° N, 9.915° Vésubie, 2002.05.15, samples No 121, 123, 126 [44.101° E, 250 m]. – Kazakhstan: Saur Mountains, 2001. N, 7.235° E, 1611 m]; Saint-Martin-Vésubie, 2012.08.15, 07.28, sample No 272 [47.055° N, 84.924° E, 2000 m]; samples No 1 - 4 [44.104° N, 7.230° E, 1766 m]; Saint- Saur Mountains: Matagul, 2001.07.28, samples No 123, 125, Martin-Vésubie, 2012.08.15, samples No 5 - 12 [44.103° 126 [47.049° N, 84.914° E, 1655 m]; Tarbagatai Mts., N, 7.233° E, 1686 m]. – Spain: Espot-1 km W, 1991. 2001.08.05, samples No 306, 201 [47.059° N, 82.368° E, 05.06, sample No 131 [42.588° N, 1.071° E, 1400 m]. 2000 m]; Tarbagatai Mts.: Sarymobe, 2001.08.05, sample No 202a [47.130° N, 82.372° E, 1887 m]; Tarbagatai Mts.: Myrmica scabrinodis NYLANDER, 1846 Sarymobe, 2001.08.05, sample No 208 [47.132° N, 82.377° 89 nest samples with 275 workers were investigated by E, 1886 m]. – Kyrgyzstan: Sary-Tschelek: Tos-Kol, CLM only and 83 nest samples with 253 workers by both 1998.07.22 [41.867° N, 71.983° E, 1600 m]. – Po-

172 land: Izbica: T-Cross, 2006.08.22, samples No 78, 68 Myrmica scabrinodis Nyl. v. rugulosoides For. type" [Forels [54.658° N, 17.419° E, 2 m]. – Russia: Odinsk-4 km handwriting], all MHN Genéve. WSW, 2008.07.13, sample No 3 [52.400°N, 103.700°E, Myrmica pilosiscapus BONDROIT, 1920: Lectotype 466 m]. – Spain: Soria, 1987.05.27 [41.760° N, 2.460° worker labeled "Mouthe Jura, nid dans Sphagnum tres hu- W, 1100 m]. – Sweden: Krokstrand, 1986 [59.000° mide, M.pilosiscapus type Bondr." [all three labels in Bon- N, 11.433° E, 60 m]. – Switzerland: GR: Val Müst- droits handwriting], "LECTOTYPE (upper specimen) desig. air, 2011.08.14 [46.633° N, 10.335°E, 1890 m]; Roche: Radchenko & Elmes 2000" [published by RADCHENKO & bogs, lectotype of M. scabrinodis var. rugulosoides [46.370° ELMES 2010]; paralectotypes: 1 worker, 2 gynes and 2 males N, 6.920° E, 380 m]; Sur-1 km S: Mähwiese, 2010.07.15, on the same pin with lectotype; 1 paralectotype worker to- samples No 17, 26 [46.514° N, 9.631° E, 1635 m]; Sala- gether with 4 males on another pin labeled "Mouthe Jura, tegnas, 2010.07.12, samples No 31, 35, 37 [46.516° N, Type, M.pilosiscapus type Bondr." [all three labels in Bon- 9.651° E, 1960 m]. – Turkey: Ardesen, 2000.08.15, droits handwriting], "PARALECTOTYPE desig. Radchen- samples No TR123, TR132, TR127, TR129 [41.230° N, ko & Elmes 2000"; IRSNB Bruxelles. 42.440° E, 1800 m]; Göle-16SW, 1986.06.16 [40.698° N, Myrmica sabuleti var. spinosior SANTSCHI, 1931: Lec- 42.464° E, 1600 m]; Posof-5 kmS, 2012.07.17 - 297 [41.473° totype worker labeled "Typus, M.sabuleti v. spinosior Sant, N, 42.741° E, 1900 m]; Keskin-3 km SE, 2012.07.25-308 Pyren. occ. Irun 25 - 26 3. 1926 Lindberg, LECTOTYPE [39.642° N, 33.624° E, 1120 m]. – Serbia: Jaile (=Jaj- desig. Radchenko & Elmes 2002" [published by SEIFERT ce?), 1985.07.03 [putatively 44.330° N, 17.270° E, 1100 m]. 2005], NHM Basel. 2 paralectotype workers on one pin, Myrmica spinosior SANTSCHI, 1931 labeled "Pyren. occ. Irun 25 - 26 3. 1926 Lindberg, M. sabuleti v. spinosior Sant, Paralectotype M. spinosior Sant- 30 nest samples with 83 workers were studied by CLM: schi det. Seifert 2004", NHM Basel. These specimens are France: Briancon, 1955.05 [44.90° N, 6.63° E, 1300 m]; in morphology and locality label consistent with the lecto- Pyr.Or.: La Tour de Carol, 1966.06 [42.470° N, 1.880° E, type. 1400 m]; Val d'Oueil, 1929.08.20, lectotype of M. reticu- Myrmica rolandi var. reticulata STÄRCKE, 1942: The lata [42.804° N, 0.503° E, 1200 m]; Tourettes-sur-Loup, lectotype sample of 8 workers has been mounted by Sant- 1955.05, samples No 1, 3 [43.720° N, 7.060° E, 400 m]; schi on a single pin in a way not allowing a reasonable in- Corsica: Zonza, 2009.04.15-b [41.733° N, 9.167° E, 780 m]. vestigation. Accordingly, the 4 best conserved specimens – Italy: Elba: Cavo (Aduorris), 1978.06 [42.860° N, were washed off, cleaned and remounted on a single pin 10.430° E, 6 m]; Firence-70 km NW: Barga, 1997.08.20 labeled "Mt. Espigno. Val d' Oueil Ht. Garon. 1200 m Ker- [44.070° N, 10.480° E, 700 m]; Umbria: Montefalco 1993. ville. 20 VIII 29 [Santschi's handwriting], M. scabrinodis 06.20 / 30 [42.880° N, 12.650° E, 440 m]. – Spain: rolandi Bond. reticulata San. [Santschi's handwriting], Lec- Aitana, 1984.03.23 [38.650° N, 0.270° W, 440 m]; Canfranc, totype + paralectotypes Myrmica scabrinodis rolandi reti- 1966.06 [42.720° N, 0.530° W, 1100 m]; Castellon, 1978 culata Santschi, des. and remounted by B. Seifert 2012" [40.0° N, 0.0° E, 40 m]; Castellon, 1985.05 [40.0° N, 0.0° [lectotype by present designation: top specimen with CW E, 40 m]; Castellon: Chodos, 1991.05.07, samples No 57, = 1.237 mm and FL / FR 1.377 ], "SYNTYPE desig. Rad- 42, 109 [40.253° N, 0.299° W, 1250 m]; Cuenca, 1983.05 chenko & Elmes 2002"; 4 workers, placed on another pin, [40.000° N, 2.000° W, 1000 m]; Cuenca, 1985.05.05 [40.0° remained in Santschi's original mounting and got laser- N, 2.0° W, 1000 m]; Espot-1 km W, 1991.05.16, sample printed labels "label copy by B. Seifert, read as: Mt. Es- No 38 [42.588° N, 1.071° E, 1700 m]; Gerona, L'Estartit- pigno. Val d' Oueil Ht. Garon. 1200 m Kerville. 20 VIII 29, 1985.05, sample No 1 [42.530° N, 3.194° E, 9 m]; Irun- Paralectotypes Myrmica scabrinodis rolandi reticulata Sant- 1926.03, lectotype of M. spinosior [43.330° N, 1.790° schi, des. B. Seifert 2012"; 3 males, 1 gyne, labeled "Val de W, 250 m]; La Seu de Urgell, 1991.05.15, sample No 71 la Frèche fond val de la Pique Hte Gar. 1600 m G.Ker- [42.379° N, 1.231° E, 1900 m]; La Seu de Urgell, 1991. ville 23 IX 30 [Santschi's handwriting], Vol nuptial mixte 05.15, sample No 158 [42.366° N, 1.262° E, 1400 m]; [Santschi's handwriting], SYNTYPE desig. Radchenko & Leon: Molinaferrera Sp 567 [42.400° N, 6.370° W, 1100 m]; Elmes 2002"; 3 males, 1 gyne, labeled only "Vol nuptial Leon: Redipollos, 1989.05 [43.000° N, 5.260° E, 1150 m]; mixte" [Santschi's handwriting] and "M. scabrinodis rolan- M. Universali: Frias: Albarracin [40.340° N, 1.620° W, di" [Santschi's handwriting]; all material in NHM Basel. 1400 m]; Puerto de Navacerrada, 1991.05.14, sample No Comments: In his original description, Santschi gave 123 [40.853° N, 4.027° W, 1250 m]; Sella Alicante, 1984 the following information on type localities: "Pyrénées cen- [38.610° N, 0.270° W, 400 m]; Teruel, 1983.05 [40.350° N, trales, Cirque d'Espingno, entre 1850 et 2000 m., août 1929 1.100° W, 1000 m]. w, g, m (types) – Val Astos, Commune d' Oô, entre 1100 Type material investigated et 1200 m., août – Val de la Frèche, fond de la vallée de la Pique, entre 1550 et 1650m. Haute-Garonne, sous des pier- Myrmica scabrinodis var. rugulosoides FOREL, 1915: res (H. Gadeau de Kerville leg.)". To explain geography, Lectotype worker together with a paralectotype worker on Haute Garonne is a region around Toulouse including the the same pin labeled "Typus", "LECTOTYPE (bottom spe- Central Pyrenees, Val de Oueil (Val Astos) is a valley in the cimen) desig. Radchenko & Elmes 2000" [published by Pyrenees approximately at 42.83° N, 0.54° E, Commune RADCHENKO & ELMES 2010], "Tourbières de Roche" [Fo- d'Oô is a village in the Pyrenees at 42.796° N, 0.506° E, rels handwriting], "Myrmica scabrinodis Nyl. v. ruguloso- 966 m and Cirque d'Espingno (or "Espingo") is situated at ides For. type" [Forels handwriting], "ANTWEB CAS- 42.728° N, 0.499° E and 1920 m. ENT0907654"; 3 paralectotype workers on a single pin RADCHENKO & ELMES (2010) published the following labeled "Typus, PARALECTOTYPE desig. Radchenko & lectotype designation: "male (upper specimen on the pin Elmes 2000, Tourbières de Roche [Forels handwriting], with 3 males and 1 queen), M. scabrinodis rolandi Bondr.

173 var. reticulata Sant [very hardly readable text, most probab- of excision superimpose. This is usually given after ly as follows] Mo Espigno, Val d'Oueu, Ha Garon., 1200 m, tilting the head by ± 45° from dorsal towards the Kerville 20.VIII.29." frontal viewing position (dorsofrontal view). The This lectotype could not be identified by an existing assumed surface is a compromise between valleys label. Santschi's original labeling cited by Radchenko and and peaks of sculpture. Elmes is exactly that of the 8 workers designated by us as CS Cephalic size; arithmetic mean of CL and CW, used lectotype and paralectotypes. Apparently, Radchenko and El- as a less variable indicator of body size. mes recorded the labels correctly but confused which castes CW Maximum cephalic width; in Myrmica this is always were associated with which labels. The statements in Sant- across the eyes. schi's original description also disagree with the situation EYE Eye-size: arithmetic mean of large (EL) and small found in the collection and the text on the lectotype label diameter (EW) of the elliptic compound eye. is a disagreement in itself: Mt. Espingo is at 1920 and not FL Maximum anterior divergence of frontal carinae at 1200 m. The smallest mismatch between Santschi's state- (= maximum frontal lobe width). In specimens with ments and the labeling is found in the worker series in frontal carinae parallel or converging frontad FL is which at least the name of location, elevation and month not defined, then FL = FR. of collection agree. Accordingly, we selected and labeled FR Minimum distance between frontal carinae. In spe- as lectotype a worker from the nest sample from Val cimens with parallel frontal carinae or ones converg- d'Oueil collected at an elevation of 1200 m. Furthermore, ing frontad FR is not defined; FR then is measured selecting a male as primary type specimen in Myrmica at the level of the centre of frontal triangle. would be a wrong decision because the delimitation of MetL Height of metapleuron including propodeal lobe closely related species is usually most difficult and in many measured in lateral view perpendicular to straight cases unsolved just in this caste. section of metapleuro-coxal border. The lower end- point of measuring line is the metapleuro-coxal bor- Methods der and the upper one the upper margin of propo- An average of three mounted workers per sample was in- deal lobe. The level of the measuring line is posi- vestigated by two-dimensional geometric morphometrics tioned in the middle between the frontalmost point and conventional linear morphometrics. The number of of subspinal excavation and the caudalmost point of investigated workers was increased to six in problematic propodeal lobe (see SEIFERT & al. 2009: fig. 1). samples. MetSp Height of subspinal excavation from upper margin Geometric morphometrics: The methodology of geo- of propodeal lobe to lower spine margin measured metric morphometrics, including optical equipment, digitiz- along dorsal continuation of measuring line for MetL. ing, symmetrizing, Generalized Procrustes Analysis, calcu- PEH Maximum petiole height measured perpendicular lating deformation grids and the used software packages are to a reference line defined as follows: The frontal described elsewhere (BAGHERIAN YAZDI & al. 2012). We endpoint of the reference line is marked by the cen- fixed 41 landmarks and 252 semilandmarks in the clype- tre of the petiole-propodeal junction and the cau- us, head capsule, mesosoma and petiole aspects and ex- dal endpoint by the centre of petiole-postpetiolar tracted a total of 316 relative warps. MorphoJ software junction (see SEIFERT & al. 2009: fig. 2). version 1.05a (KLINGENBERG 2011) was used to calculate PEL Maximum measurable diagonal petiole length from and to display the shape changes related to the LDA scores tip of subpetiolar process to dorsocaudal corner of as wireframe graphs. The basic requirements for geomet- caudal cylinder. (Do not confuse this with the cor- ric analysis were fulfilled in the investigated material: ner of the movable inner sclerite.) a) Variation of the specimens in shape space was per- PEW Maximum width of petiole. fectly correlated with tangent space for all anatomical as- PoOc Postocular distance. Use a cross-scaled ocular pects. micrometer and adjust the head to the measuring b) Centroid sizes of both species were normally distri- position of CL. Caudal measuring point: median buted in all anatomical aspects (Asymp. Sig. > 0.05) and occipital margin; frontal measuring point: median means of centroid sizes were significantly different be- head at the level of the posterior eye margin. Note tween the species in all anatomical aspects (p < 0.001). that many heads are asymmetric and average the c) No directional asymmetry was demonstrable in the left and right postocular distance (see SEIFERT & investigated ants with any test system applied. al. 2009: fig. 3). Conventional linear morphometrics: The optical PPHL Length of longest hair on dorsal postpetiole. equipment used, the character recording methods and esti- PPW Maximum width of postpetiole. mation of measuring errors are given in SEIFERT (2011). SL Maximum straight line scape length. Distal measur- The definitions for 19 characters are as follows. ing point: most distal point of dorsal lamella of CL Maximum cephalic length in median line; the head hinge joint capsula. Proximal measuring point: most must be carefully tilted to the position with the true proximal point of scape shaft near neck of articu- maximum. Excavations of posterior head margin lar condyle. Note that the border region between and / or clypeus reduce CL. Longitudinal carinae or shaft and condylar neck is usually asymmetric. To rugae on anterior clypeus are included into the mea- measure the real maximum, avoid positions near to surement – if exactly median, in their full height SVP C; instead use positions between the SVPs F and, if of doubtful position, in their half height. and D! In species with basal scape lobes or dents ClyEx Depth of excision on anteromedian clypeal margin (e.g., M. schencki, M. scabrinodis and M. lobicornis in a position in which dorsal and ventral margins group) the lobes are excluded from measurement!

174 Tab. 1: Worker nest sample means of RAV-corrected data of Conventional Linear Morphometrics given as arithmetic mean ± standard deviation [lower extreme, upper extreme]. F values and significancy levels p are from an univariate ANOVA; the F values of the most separating characters are given in heavy type. M. martini ANOVA M. scabrinodis ANOVA M. spinosior (n = 23) F; p (n = 89) F; p (n = 30) CL / CW (1150) 1.020 ± 0.017 14.81 1.032 ± 0.014 n.s. 1.035 ± 0.019 [0.986, 1.055] 0.0001 [0.983, 1.064] [0.999, 1.082] ClyEx / CS) [%] 0.68 ± 0.20 n.s. 0.76 ± 0.19 7.76 0.62 ± 0.36 (1150) [0.18,1.17] [0.12, 1.20] 0.006 [0.00, 1.27] CS [µm] 1089 ± 45 n.s. 1079 ± 57 95.44 1200 ± 64 [1027, 1175] [953, 1198] 0.0001 [1083, 1342] EYE (1150) 0.202 ± 0.003 n.s. 0.201 ± 0.005 81.72 0.192 ± 0.005 [0.196, 0.208] [0.189, 0.213] 0.0001 [0.184, 0.207] FL / CS (1150) 0.427 ± 0.008 69.84 0.449 ± 0.012 21.83 0.438 ± 0.011 [0.412, 0.444] 0.0001 [0.425, 0.493] 0.0001 [0.413, 0.459] FL / FR (1150) 1.287 ± 0.040 83.37 1.415 ± 0.064 30.48 1.340 ± 0.067 [1.189, 1.372] 0.0001 [1.317, 1.589] 0.0001 [1.239, 1.469] FR / CS (1150) 0.332 ± 0.009 30.80 0.317 ± 0.012 16.742 0.328 ± 0.013 [0.312, 0.351] 0.0001 [0.286, 0.352] 0.0001 [0.307, 0.351] MetL / CS (1150) 0.237 ± 0.006 4.32 0.240 ± 0.007 59.12 0.228 ± 0.009 [0.222, 0.248] 0.04 [0.225, 0.255] 0.0001 [0.202, 0.246] MetSp / CS (1150) 0.159 ± 0.005 7.49 0.163 ± 0.008 226.91 0.190 ± 0.010 [0.150, 0.170] 0.007 [0.143, 0.187] 0.0001 [0.174, 0.207] MetSp / MetL (1150) 0.670 ± 0.027 n.s. 0.681 ± 0.031 272.3 0.835 ± 0.062 [0.622, 0.723] [0.579, 0.756] 0.0001 [0.711, 1.004] PEH / CS (1150) 0.332 ± 0.008 n.s. 0.333 ± 0.009 n.s. 0.329 ± 0.011 [0.309, 0.347] [0.313, 0.358] [0.314, 0.363] PEL / CS (1150) 0.476 ± 0.008 n.s. 0.473 ± 0.013 36.52 0.489 ± 0.013 [0.463, 0.494] [0.444, 0.511] 0.0001 [0.452, 0.523] PEW / CS (1150) 0.271 ± 0.009 n.s. 0.276 ± 0.010 n.s. 0.276 ± 0.013 [0.258, 0.292] [0.254, 0.296] [0.250, 0.314] PoOc / CL (1150) 0.421 ± 0.007 17.06 0.428 ± 0.008 10.13 0.433 ± 0.007 [0.408, 0.435] 0.0001 [0.413, 0.449] 0.02 [0.419, 0.448] PPHL / CS (1150) 0.175 ± 0.006 39.18 0.164 ± 0.008 46.60 0.176 ± 0.009 [0.158, 0.186] 0.0001 [0.146, 0.180] 0.0001 [0.160, 0.196] PPW / CS (1150) 0.405 ± 0.012 n.s. 0.405 ± 0.013 16.88 0.394 ± 0.013 [0.379, 0.431] [0.378, 0.438] 0.0001 [0.367, 0.427] SL / CS (1150) 0.776 ± 0.010 16.08 0.787 ± 0.012 197.06 0.828 ± 0.017 [0.755, 0.799] 0.0001 [0.763, 0.814] 0.0001 [0.800, 0.861] SP / CS (1150) 0.378 ± 0.021 10.08 0.395 ± 0.022 15.17 0.376 ± 0.024 [0.339, 0.419] 0.002 [0.335, 0.443] 0.001 [0.337, 0.426] SW / SL (1150) 0.141 ± 0.009 26.22 0.153 ± 0.010 n.s. 0.154 ± 0.012 [0.116, 0.154] 0.0001 [0.132, 0.177] [0.136, 0.187] tan α (1150) 0.949 ± 0.119 6.33 0.865 ± 0.142 175.02 0.463 ± 0.152 [0.73,1.09] 0.013 [0.44,1.14] 0.0001 [0.13,0.91]

SP Maximum length of propodeal spines as bilateral ment (see SEIFERT & al. 2009: fig. 6). Dorsal view arithmetic mean. Measured in dorsofrontal view from (SVP D) is directed perpendicular to this moving spine tip to a point at bottom of interspinal meniscus plane (in this position the anterior margins of upper (see SEIFERT & al. 2009: fig. 4). With the spines' and lower lobe of the distal scape end are congru- dorsal edge in measuring plane, the spine tip must ent and the basal curvature of scape is not or only be focused at a magnification with low depth of weakly visible). Frontal view (SVP F) and caudal focus. Then, while keeping this focusing, the sharp- view (SVP C) are within the moving plane and per- est point at the bottom of interspinal meniscus is the pendicular to the longitudinal scape axis – i.e., when basal measuring point. This mode of measuring is the scape is imagined to be directed strictly laterad less ambiguous than other methods but results in from head, SVP F is the frontal and SVP C the some spine length in species with reduced spines. caudal aspect of scape. SVPs such as CD and DF SVP Standard viewing positions of scape defined by describe intermediate viewing positions. SVP L is their position relative to the moving plane of the the view along the longitudinal axis of scape from hinge joint between scape and first funiculus seg- its distal to proximal end.

175 SW Worker and gyne: maximum distance between cau- The applied species concept: We applied the species dal margin of basal scape lobe and anterior scape concept of DE QUEIROZ (2007). As delimitation criterion margin measured perpendicular to longitudinal scape we used phenotypic distinctness. The decision threshold axis. The scape must be carefully tilted until the in favour of heterospecificity was that the discriminant maximum distance is situated in the measuring plane. analysis controlling the explorative data analysis must not This is given in approximately SVP DF. In cases place more than 2% of the classified samples in the inter- when maximum SW would be measured proximal specific overlap range. of basal break point of anterior scape margin, a point on an imagined linear continuation of the an- Results and Discussion terior scape margin distal of the break is used as an- Myrmica martini sp.n. terior measuring point. tan α The tangens of the angle alpha under which the ba- Etymology: Named after the titular Saint of the locality sal scape lobe slopes caudad relative to the plane Saint-Martin-Vésubie situated close to the locus typicus. in which SVP L or SVP C are running. The angle Type material: Holotype worker labeled "FRA: is measured in SVP L in the following way. Fo- 44.1002° N, 7.2332° E St.-Martin-Vésubie-3.8NNW 1629 cusing the distal scape end, the horizontal axis of the m, Larix-Pinus, clearing, Schultz 2002.05.15 – 126" and cross-scaled ocular micrometer (x-axis) is aligned "Holotype Myrmica martini Seifert & al."; 8 worker para- with the plane of SVP L and C. Then, after focus- types on three other pins and 120 worker paratypes in etha- ing down to the level of basal scape lobe without nol with identical locality labels and "Paratype Myrmica changing the former adjustment, the y- and x-values martini Seifert & al."; all material stored in Senckenberg (or opposite and adjacent legs) of tangens α can Museum of Natural History Görlitz. be read in the cross-scale ocular micrometer. This is Description: Worker (Figs. 1 - 6, Tab. 1, all morpho- done in both scapes and needs some training. See metric ratios given in the following verbal description are also figures in SEIFERT (2007) and SEIFERT & al. arithmetic nest sample means of primary data – i.e., without (2009). removal of allometric variance): Myrmica martini sp.n. Removal of allometric variance: In order to make seems to lack any exposed morphological character and is shape components such as CL / CW, SL / CS or SP / CS most similar to M. scabrinodis. Medium-sized (CS 1089 µm). interspecifically comparable in synoptic tables independent Head with a straight posterior margin and strongly convex from body size (Tab. 1), a removal of allometric variance sides and not elongated (CL / CW 1.023), postocular dis- (RAV) was performed with the procedure described by SEI- tance rather low (PoOc / CL 0.420). Frontal lobes mode- FERT (2008). RAV was calculated for the assumption of all rately diverging (FL / CS 0.426) and frontal width com- individuals having an identical cephalic size of 1.15 mm. parably large (FR / CS 0.331) in terms of related species, Overall genus-specific RAV functions were applied the frontal carinae reaching caudad only to level of eyes. parameters of which were calculated as the arithmetic mean Eyes with few microsetae and medium-sized (EYE / CS of the species-specific functions of 36 Palaearctic Myrmica 0.200). Anterior clypeal margin in dorsofrontal view not or species with sufficient sample size. It can be seen from only feebly emarginated (ClyEX / CS 0.71%). Scape mode- the functions below that allometries are rather weak and rately long (SL / CS 0.781), with a clearly developed dor- usually less than 5% per 400 µm CS change (this is the sal and caudal carina at base – the plane demarcated by average intraspecific size difference between the smallest these carinae form a caudoventral slope by an angle of more and largest worker). However, SP / CS grows by 10.4% or less 44° (tan α 0.949) and is not very wide (SW / SL and ClyEx / CS by 27% from smallest to largest workers. 0.140), scape base in caudal view varying from curved The RAV functions were as follows (Fig. 3a) to almost angular. The convexity of dorsal meso- CL / CW1.15 = CL / CW / (-0.0487 * CS + 1.0900) * 1.0340 somal profile is interrupted by a rather deep metanotal de- SL / CS 1.15 = SL / CS / (-0.0802 * CS + 0.8947) * 0.8024 pression. Propodeal spines acute, moderately long (SP / EYE / CS1.15 = EYE / CS / (0.0179 * CS + 0.1772) * 0.1978 CS 0.372), their axes in dorsal view only diverging by 30 - FL / CS 1.15 = FL / CS / (0.0227 * CS + 0.4036) * 0.4297 34°, in lateral view weakly erected, deviating from longi- FR / CS1.15 = FR / CS / (0.0095 * CS + 0.3222) * 0.3331 tudinal mesosomal axis by 25 - 30°. Central height of pro- PEW / CS 1.15 = PEW / CS / (-0.0201 * CS + 0.2842) * 0.2610 podeal lobe clearly larger than equal-level height of sub- PPW / CS 1.15 = PPW / CS / (0.0709 * CS + 0.3208) * 0.4024 spinal excavation (MetL / CS 0.237, MetSp 0.159). Peti- PEH / CS 1.15 = PEH / CS / ( 0.0051 * CS + 0.3230) * 0.3288 ole in lateral view with almost straight dorsal profile that PEL / CS 1.15 = PEL / CS / (-0.0259 * CS + 0.4975) * 0.4677 slopes caudad with only a suggested step, dorsal and fron- PPHL / CS 1.15 = PPHL / CS / (-0.0679 * CS + 0.2535) * 0.1754 tal profile of petiole node form an angle of 90°. Petiole in SP / CS 1.15 = SP / CS / (0.0876 * CS + 0.2353) * 0.3360 dorsal view with weakly convex sides, its width about 67% MetL / CS1.15 = MetL / CS / (-0.0126 * CS + 0.2487) * 0.2342 of postpetiolar width. Setae are present on all dorsal parts MetSp / CS1.15 = MetSp / CS / (-0.0113 * CS + 0.1880) * 0.1749 of body, those on dorsum of postpetiole are moderately PoOc / CL1.15 = PoOc / CL / (0.0098 * CS + 0.4154) * 0.4266 long (PPHL / CS 0.180). Head including clypeus and meso- SW / SL1.15 = SW / SL / (0.0109 * CS + 0.1385) * 0.1511 soma with rather strong longitudinal rugosity, about 12 - tan α1.15 = tan α / (0.391 * CS + 0.199) * 0.649 14 rather linear rugae are found between the most approx- ClyEx / CS1.15 = ClyEx / CS / (-0.460 * CS + 1.193) * 0.664 imated parts of frontal carinae. Whole body usually rather FL / FR1.15 = FL / FR / (-0.0572 * CS + 1.4270) * 1.3612 uniformly medium brown with a weak yellowish compo- All linear discriminant analyses were run with the SPSS nent and sometimes with a lighter mesosoma. Ecology and 16.0 software package. distribution: found in submontane to subalpine grassland

176

Figs. 1 - 3: Myrmica martini sp.n.: (1) head of holotype; (2) mesosoma and waist of holotype in lateral view; (3a) scape of holotype in the standard viewing po- sition D; (3b) scape of a paratype (from the holotype nest) in the standard view- ing position C.

Figs. 4 - 6: Wireframe graphs of Myrmica martini sp.n. (blue lines) and of M. scabrinodis (red lines) with interspecific differences exaggerated by the factor three. (4) Head shape; (5) petiole shape; (6) mesosoma shape. or grassy woodland clearings of the Pyrenees and French ferences: Myrmica martini sp.n. has a significantly smaller Alps (Fig. 7). Nests are under stones or in grass tussocks. frontal lobe distance, a relatively wider head, a larger mini- Biology: unknown. mum frons width, a shorter postocular distance and shorter propodeal spines. All these differences are also indicated by Differential diagnosis derived from geometric and CLM (Tab. 1). In lateral aspect of petiole, M. martini sp.n. conventional morphometrics shows a shorter petiole node relative to petiole height, the Both geometric morphometrics (GM) and conventional lin- angularity between the dorsal and anterior profile is less ear morphometrics (CLM) showed that Myrmica scabri- distinct, the dorsal plane of petiole is shorter and slopes nodis and M. martini sp.n. are extremely similar. The wire down to caudal cylinder without a distinct step. Further- frame graphs exaggerated by a factor of three (Figs. 4 - 6) more, M. martini sp.n. has significantly longer setae on revealed only very few and often minute interspecific dif- dorsal postpetiole (Tab. 1). The shape of basal scape in

177

Fig. 7: Sampling sites of Myrmica martini sp.n. (blue dots) and of M. scabrinodis (red dots) in the Euro- caucasian range.

Tab. 2: Three diagnostic characters to distinguish taxa of the Myrmica scabrinodis and M. sabuleti species complex given as mean of the holo/lectotype samples – the profile of petiole, the sloping angle α of basal scape lobe and the ratio be- tween height of subspinal excavation and height of metapleural lobe.

Holo/lectotype sample Dorsum of petiole node tan α MetSp/MetL M. martini sp.n. truncate large: 0.955 small: 0.670 M. s. var. rugulosoides truncate large: 0.668 small: 0.699 M. pilosiscapus truncate large: 0.787 small: 0.689 M. spinosior slightly convex small: 0.427 large: 0.855 M. r. var. reticulata slightly convex small: 0.387 large: 0.953

SVP C is no reliable discriminator from M. scabrinodis be- a posterior probability of belonging to this cluster of p > cause this character varies in M. martini from nearly curved 0.970 with the lectotype samples of M. spinosior and M. (Fig. 3a) to almost angular (comparable to the situation in reticulata showing p = 0.9994 and p = 1.0000 respectively. M. scabrinodis). In the sections below we present conclu- All M. martini and M. scabrinodis samples had a posterior sive data-based evidence that M. martini sp.n. is no syno- probability of belonging to the M. spinosior cluster of p < nym of five morphologically similar taxa. 0.032 with the lectotype samples of M. scabrinodis var. rugulosoides and M. pilosiscapus showing p = 0.0000 and Demonstration of distinctness of Myrmica spinosior p = 0.0316. Figure 8 shows some 2% of doubtful alloca- from members of the M. scabrinodis complex tions between M. martini and M. scabrinodis with this char- Myrmica spinosior SANTSCHI, 1931 and its junior synonym acter system and type of analysis. This problem is treated M. rolandi var. reticulata STÄRCKE, 1942 (first available in the next section by more advanced geometric analyses. use of M. scabrinodis rolandi var. reticulata SANTSCHI, A full separation of M. spinosior from M. martini and M. 1931) have been described from the Pyrenees and we show scabrinodis is also provided by a plot of the 1st and 2nd in the following that both are not synonyms of M. martini components of a PCA (not shown). Accordingly, the sepa- sp.n. Myrmica spinosior, M. sabuleti MEINERT, 1861 and rate identity of M. spinosior is clearly confirmed by both M. lonae FINZI, 1926 together form the M. sabuleti spe- cies complex (SEIFERT 2005). In the normal situation, the workers of the M. sabuleti complex can be subjectively separated from M. martini sp.n. and M. scabrinodis by a higher and not truncate petiolar node, a smaller sloping angle of basal scape lobe α and a higher subspinal exca- vation (MetSp) relative to the height of metapleural lobe (MetL). These differences are also expressed in the lecto- type samples of M. spinosior, M. rolandi var. reticulata, M. scabrinodis var. rugulosoides and M. pilosiscapus (Tab. 2) but many, even experienced, observers have difficulties to recognize these useful traits. Accordingly, there is a need to apply more complex numeric approaches to demonstrate the differences in the workers convincingly. We had no data of geometric morphometrics at hand for Myrmica spinosior but there are enough samples with data of conventional linear morphometrics. A three-class LDA considering all 19 characters provided a full separa- Fig. 8: Nest sample means of a canonical variance analysis tion of the M. spinosior cluster from the M. martini and M. of Myrmica spinosior (grey triangles), M. scabrinodis (grey scabrinodis clusters (Fig. 8). All M. spinosior samples had rhombs) and M. martini sp.n. workers (white squares).

178

Fig. 9: NC-Ward clustering of the first ten relative warps of Myrmica martini sp.n. (red) and M. scabrinodis (black). Arrows point to misplaced samples which are rectified if run as wild-cards in a controlling linear discriminant analysis.

179 explorative and hypothesis-driven data analyses and we can showed the same result which is expected as the number also state here that M. r. var. reticulata is a junior syno- of cases in the smallest class (n = 84 in M. martini) was nym of M. spinosior. sixfold larger than the number of considered characters. Over all samples, the LDA classified 95.9% of the worker Separation of Myrmica martini from M. scabrinodis individuals correctly with 82.3% of individuals being clas- by geometric morphometrics and NC-clustering sified with posterior probabilities of p > 0.95. a) NC-clustering considering the first ten Relative Conventional linear morphometrics Warps: Geometric morphometrics (GM) was performed in a total of 23 nest samples with 84 worker specimens of FL / FR is clearly the most discriminative shape compo- Myrmica martini sp.n. and 83 samples with 252 workers nent of conventional linear morphometrics (Tab. 1). Yet, of M. scabrinodis. The GM analysis extracted a total of even the best character is weak: on the worker individual 316 Relative Warps (RWs). NC-clustering considering all level and without removal of allometric variance, 57.2% 316 RWs did not show a clear and reasonable structure. of specimens are found in the interspecific overlap range. Therefore, we restricted the analysis to the first ten RWs This illustrates the extreme similarity of Myrmica martini which describe 63.3% of variance in the data set. NC-Ward sp.n. and M. scabrinodis. We used the species hypothe- clustering of RWs 1 - 10 showed two clearly different main sis found by geometric morphometrics (GM) as input for branches (Fig. 9) and the controlling LDA was run with a LDA considering all 19 characters of conventional line- this hypothesis – with exception of a wild-card setting in ar morphometrics (CLM). NC Ward clustering (Fig. 11) the sample St.-Martin-Vésubie-No 2. NC-Ward positioned classified 2.8% of the samples in disagreement with the this sample within M. scabrinodis but there was a strong classification by GM (Fig. 10). The LDA confirmed these geographic counter-indication because all samples from this three classifications, however with insignificant posterior spot should belong to M. martini. The two-class LDA probabilities: St.-Martin-Vésubie No 2 was classified as changed the classification of St.-Martin-Vésubie-No 2 to M. scabrinodis (p = 0.556) whereas Tourettes-sur-Loup M. martini but also that of the sample Tourettes-sur-Loup 1955.06 and Sur-2010.07.15 No 17 were changed to M. 1955.06 to M. scabrinodis. As a consequence, the cluster- martini (p = 0.821 and p = 0.632). The type samples of ing error of NC-Ward on the two-class level was 1.9% if M. pilosiscapus and M. s. rugulosoides were allocated to the LDA indication is considered as the deciding system. the M. scabrinodis cluster with p = 0.988 and p = 0.671. The lectotype series of M. scabrinodis var. rugulosoides Over all samples and after accepting those three changes, and of M. pilosiscapus were clearly allocated by the LDA the LDA classified 94.2% of the worker individuals cor- to M. scabrinodis with the sample means of posterior pro- rectly with 71.0% of individuals being classified with pos- babilities being p = 0.960 and p = 0.973 respectively. terior probabilities > 0.95. NC-Ward clustering suggested the existence of three A stepwise LDA reducing the number of characters to groups – Myrmica martini and two entities within M. sca- seven (SL / CS, FL / CS, FR / CS, PPHL / CS, PoOc / CL, brinodis. Accordingly, we ran the controlling method NC- SW / SL, FL / FR) did not basically change the situation: K-means clustering assuming three groups. This run showed The posterior probabilities were p = 0.671 in St.-Martin- an error of 0% against the two-class LDA classification: Vésubie No 2 , p = 0.863 in Tourettes-sur-Loup 1955.06 no M. martini sample was allocated to one of the two M. and p = 0.671 in Sur-2010.07.15 No 17 whereas the type scabrinodis clusters and, conversely, no M. scabrinodis samples of Myrmica pilosiscapus and M. s. rugulosoides sample to M. martini. Hence, the agreement of NC-K-means were allocated to the M. scabrinodis cluster with p = 0.983 and NC-Ward was 98.1% on the two-class level. and p = 0.622. Also using the reduced character set, NC- b) NC-clustering considering the most diagnostic Ward performed the same classification changes described Relative Warps: Taking the hypotheses formed above and in the previous section (Fig. 11) while NC-K-means in order to possibly improve the success of NC-clustering, repeated these three changes and added another one: we ran a stepwise LDA considering all 316 RWs. This re- St.-Martin-Vésubie No 10 was classified as M. scabri- sulted in a reduction to 14 RWs: 1, 2, 4, 7 - 10, 12, 14, 18, nodis. 22, 23, 25 and 28. Considering these 14 characters, NC- Ward clustering (Fig. 10) and the controlling LDA agreed The type series of Myrmica scabrinodis NYLANDER is by 100% – even the samples St.-Martin-Vésubie-No 2 and clearly different from M. martini sp.n. Tourettes-sur-Loup 1955.06 were correctly placed by the The type series of Myrmica scabrinodis NYLANDER, 1846 explorative data analysis. The lectotype series of Myrmica – consisting of three workers labeled "Kuusamo, W.Nyland. scabrinodis var. rugulosoides and of M. pilosiscapus were and Mus. Fenn." – was not ordered from Helsinki for this clearly allocated by the LDA to M. scabrinodis with the study because there was no danger of a synonymy with nest means of posterior probabilities being p = 0.996 and M. martini sp.n. In 1995, the senior author studied this p = 0.999. series by subjective eye-inspection and performed a few Similar to the analysis with RWs 1 - 10, NC-Ward clus- simple measurements in one specimen. In overall impres- tering suggested the existence of three groups and we ran sion, all three type specimens fully matched the typical M. NC-K-means clustering under this assumption. The run scabrinodis phenotype. Furthermore, FL / FR was 1.465 showed an error of 0% against the two-class LDA classi- in the single measured specimen which is clearly above fication: no Myrmica martini sample was allocated to one the range of variation known in 84 workers of M. martini: of the two M. scabrinodis clusters and, conversely, no M. 1.291 ± 0.048 [1.167, 1.397] . These data are normally dis- scabrinodis sample to M. martini. Accordingly, the agree- tributed and there is a probability of p < 0.00015 for M. ment of NC-K-means, NC-Ward and LDA was 100% on martini to achieve a FL / FR value of 1.465 or larger. The the two-class level. A leave-one-out-cross-validation LDA third strong argument against a synonymy is the distance

180

Fig. 10: NC-Ward clustering of the 14 most diagnostic relative warps separating Myrmica martini sp.n. (red) and M. scabrinodis (black). This classification is fully confirmed by NC-K-means and NC-NMDS-K-means clustering and the controlling linear discriminant analysis.

181

Fig. 11: NC-Ward clustering of the seven most diagnostic characters of conventional linear morphometrics to separate Myrmica martini sp.n. (red) and M. scabrinodis (black). Arrows point to samples positioned in disagreement with the clustering based on geometric morphometrics (Fig. 9).

182 of 2700 km between the boreal type locality of M. scabri- rent state of imaging technology and software development. nodis and the next known site of M. martini sp.n. This means more than a threefold processing time, summing up to eight hours for a standard sample of three workers, A simpler way to discriminate M. martini from M. and prevents the present application of GM in routine in- scabrinodis vestigations of thousands of samples. However, 165 min- The data presented above show an extreme similarity of utes required from specimen preparation to the final com- the two species. Accordingly an easy way of determina- putation result is very little compared to the complete pro- tion is not possible. We tried to simplify the species deli- cessing time of other advanced methods in taxonomy such mitation procedure by using absolute linear measurements as microtomography or Next Generation Sequencing. This only and by reducing the number of characters for the important disadvantage is usually not communicated. condition that the error at nest sample level was < 5%. We extracted a morphometric method requiring 8 - 10 minutes Acknowledgments per specimen. With all measurements recorded in milli- We wish to thank Christophe Galkowski (Saint-Aubin de metres, a linear discriminant function Medoc) for travelling to the type locality and kindly col- D(5) = 0.466 *SL -37.204 * FL + 45.95 * FR + 42.11 *PPHL - lecting many samples of M. martini sp.n., the curators 16.92 *SW - 4.103 Daniel Burckhardt (NHM Basel), Bernhard Merz (MHN resulted in an error of 3.6% on the nest sample level while Genéve) and Wouter Dekoninck (IRSNB Bruxelles) for the number of misclassified worker individuals was 9.4%. loan of type specimens, Andreas Schulz (Leichlingen) for Nest samples or worker individuals with D (5) < 0.633 are providing material from Greece and Anatolia, Graham classified as M. scabrinodis, those above this threshold as Elmes (Wareham / Dorset) for providing material from M. martini. Within the range of D (5) < 0.186 and D(5) > Anatolia. 0.825, the classification error on the individual level falls below 2%. References Final conclusions BAGHERIAN YAZDI, A., MÜNCH, W. & SEIFERT, B. 2012: A first demonstration of interspecific hybridization in Myrmica ants Using GM, we found a clear morphological separation of by geometric morphometrics (Hymenoptera: Formicidae). – Myrmica martini sp.n. from M. scabrinodis. The indepen- Myrmecological News 17: 121-131. dent system of CLM basically confirmed this classification DE QUEIROZ, K. 2007: Species concepts and species delimitation. and disagreed by 2.8% only. For the reasons given below, – Systematic Biology 56: 879-886. we consider GM as the more powerful system and the KLINGENBERG, C.P. 2011: MorphoJ: an integrated software pack- leading indicator. The morphological classification is high- age for geometric morphometrics. – Molecular Ecology Re- ly correlated with a geographical pattern: Myrmica martini sources 11: 353-357. sp.n. was found in the submontane to subalpine zones of RADCHENKO, A.G. & ELMES, G.W. 2010: Myrmica ants (Hyme- the Pyrenees and the French Alps whereas M. scabrinodis noptera: Formicidae) of the Old World. – Natura optima dux is found in temperate and subboreal Eurasia outside this Foundation, Warsaw, 789 pp. region. The two M. scabrinodis samples from Tourettes- SEIFERT, B. 2000: FINZI, 1926 – a species sepa- sur-Loup are very close to the localities of M. martini in rate from Myrmica sabuleti MEINERT, 1861 (Hymenoptera: the French Alps. The data we have so far indicate no or Formicidae). – Abhandlungen und Berichte des Naturkunde- only minimal range overlap between the two species and museums Görlitz 72: 195-205. we consider them as parapatric species with probably sig- SEIFERT, B. 2005: Rank elevation in two European ant species: nificant reproductive barriers. It remains to be studied by Myrmica lobulicornis NYLANDER, 1857, stat.n. and Myrmica means of nuDNA markers how strong these barriers really spinosior SANTSCHI, 1931, stat.n. (Hymenoptera: Formicidae). are. The geographic distribution of M. martini strongly sug- – Myrmecologische Nachrichten 7: 1-7. gests that it survived the last glaciation in a lowland re- SEIFERT, B. 2007: Die Ameisen Mittel- und Nordeuropas. – Lutra, fuge south of the Pyrenees or between the Pyrenees and Klitten, 368 pp. French Alps. The main postglacial spreading was probably SEIFERT, B. 2008: Removal of allometric variance improves spe- a movement up into the montane and subalpine zones of cies separation in multi-character discriminant functions when these mountains but there is no information on the situation species are strongly allometric and exposes diagnostic charac- in the planar and colline regions of South and Central ters. – Myrmecological News 11: 91-105. France. SEIFERT, B. 2009: Cryptic species in ants (Hymenoptera: For- We suppose a lower performance of CLM compared micidae) revisited: We need a change in the alpha-taxonomic to GM. This is indicated by two classifications being in approach. – Myrmecological News 12: 149-166. strong disagreement with zoogeographic arguments. The SEIFERT, B. 2011: A taxonomic revision of the Eurasian Myrmica hypothesis of a lower performance is also supported by the salina species complex (Hymenoptera: Formicidae). – Soil Or- lower percentage of individual workers classified with pos- ganisms 83: 169-186. AGHERIAN terior probabilities of p > 0.95. Furthermore, B SEIFERT, B., RITZ, M. & CZŐSZ, S. 2014: Application of Explo- YAZDI & al. (2012), investigating an interspecific hybrid sce- ratory Data Analyses opens a new perspective in morphology- nario in Myrmica and running CLM and GM on identical based alpha-taxonomy of eusocial organisms. – Myrmecologi- samples in parallel, also suggested a slightly lower perform- cal News 19: 1-15. ance of CLM. However, there is also an advantage of CLM: SEIFERT, B., SCHLICK-STEINER, B.C. & STEINER, F.M. 2009: Myr- the much lower processing time. Recording 19 characters in mica constricta KARAVAJEV, 1934 – a cryptic sister species CLM needs about 50 minutes per worker specimen where- of FINZI, 1926 (Hymenoptera: Formicidae). as the whole GM procedure needs 160 minutes at the cur- – Soil Organisms 81: 53-76.

183